I had thought that after these posts [1] [2] [3] [4], Mr. Lawler had foregone further attempts at macroeconomic analysis. Alas, no. In Obama’s Job Search, Mr. Lawler states:
…It should be mentioned that the CBO hasn’t provided any evidence whatsoever that the stimulus created 2-5 million jobs. …
I do wonder what would convince Mr. Lawler that any job had been created by ARRA. In the past, when I observed that macroeconomists, including Greg Mankiw, typically analyzed macro impacts in the manner CBO had, Mr. Lawler averred:
Dr. Chinn, my article was about messaging, not economic analysis. I do not claim to be a “real macroeconomist” or anything other than a journalist. I’ll have a response to your criticism soon.
The response includes this statement:
So if you had a priori reasons to doubt the forecast the CBO made in February (and you very well might, given the track record of parts of that forecast), those reasons also apply to their analysis of the stimulus’s job creation. The CBO has not provided any further evidence that would validate or falsify the administration’s predictions, or that would confirm their “jobs created or saved” claims.
So once again, the typical non-economists’ confusion of forecasting and prediction comes to fore. [5] A recent attempt to deal with such confusions is here Worthwhile Canadian Initiative. Not that I expect it to work on those impervious to education.
Anyway, as before, I thank Mr. Lawler for a very long laugh, on lunar new year! (Allusion to a mysterious “Cindy Romer” merely adds to the fun.)
I suggest you add the following to preface your comments section:
Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data, ability to repeat discredited memes, and lack of respect for scientific knowledge. Also, be sure to create straw men and argue against things I have neither said nor even implied. Any irrelevancies you can mention will also be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.
Just give credit to The Big Picture.
“…It should be mentioned that the CBO hasn’t provided any evidence whatsoever that the stimulus created 2-5 million jobs. …”
The problem: one can never prove the accuracy of the counterfactual, because it can never be directly observed, only imputed from evidence and assumptions. In contrast, the forecaster is subject to an ex-post check on the accuracy of his art. Thus, although generally more esteemed by the profession, the predictor is infinitley harder to discredit by facts than the forecster.
Um… Menzie, are you sure that macroeconomists “typically” analyze effects in this way? Aren’t the models used by the CBO just old-school, massive, ludicrously poorly identified Keynesian models, of the kind that got tossed out by macroeconomists three decades ago? (Sims 1980, right?) I think that the stimulus probably did help, but it is reasonable to be skeptical of these models.
Oh, but why stop there. Consider:
“Also, the fact that TARP will cost taxpayers $28 billion instead of much more is almost totally irrelevant to whether it was a good idea.”
TARP is a matter of economic policy, so I would think that economic outcomes would matter, but even in other policy areas, good policy analysis includes cost/benefit analysis. Lawler is making a truly loopy claim when he says the cost of a policy is “almost totally irrelevant to whether it was a good idea.”
But let’s keep that standard in mind when reading the rest of what Lawler has to say. Do his criticisms address the question of whether Obama’s policies were a good idea? Part of the answer to that question involves whether Lawler addresses whether Obama’s policies worked.
When Lawler finds Geithner’s defense of policy uninspiring, that does not address whether policy worked or was a good idea. When Lawler demands that score-keepers provide evidence of a counterfactual, he is not directly addressing whether policy worked or was a good idea, but is instead demanding an impossible level of proof before allowing that policy was a good idea. Ignoring the obvious intellectual (and perhaps ethical) problems with Lawler’s position, let’s consider where the debate would be if proponents of Obama policies made the same demand of Lawler. What evidence is there that policy didn’t work? And remember, the evidence has to be drawn from a counterfactual situation in which Obama’s policies were not tried.
The possible outcomes pretty clearly come down to just two. Either everybody would have to shut up, having no evidence that meets the standard Lawler demands, or both sides could run around making whatever claims they want, because no claim could ever meet Lawler’s standard.
The reality is that we have a mix of two situations. One is the level-headed use of the tools that are available too us, however, imperfect, and the other – the one Lawler’s masters want to foster, just as they did in the tobacco debate and the climate change debate and the tax-cut-cure-all debate – is both sides running around making whatever claims they want, because people like Lawler are paid to cast doubt on any evidence that doesn’t suit their masters.
And while we’re at it, let’s just keep in mind Lawler’s standard for his own assertions: Journalists can say really ignorant things in really strong terms because they aren’t specialists, and so aren’t required to know what they are saying. That is a far cry from the standard he requires for economic analysis – deliver proof that can’t exist or stand down.
In a world full of fair-minded, honest people, journalists and “opinionators” would hold themselves to high standards, and specialists would be held to realistic ones. But not in Lawler’s world.
Random Student: I suspect you have not been following the debate that has spanned the months, so at the risk of repeating myself to the Econbrowser readers, let me restate: the issue is how to evaluate the impact of any policy measure, when all one observes is the outcome. One has to compare against a counterfactual of some sort. Using Mr. Lawler’s standards, one can never determine a policy’s impact because one cannot observe the counterfactual. (Note: As far as I can tell, he resists simulating a model to w/o the measure, as a counterfactual, and then simulating with the measures, and using the difference as the measure of impact).
Hence, the issue is not what type of model is used (DSGE vs. large scale macroeconometric), but how to measure the impact, and would apply with equal force to the DSGE-based projections, such as those John Taylor cited.
In fact, one might reasonably argue the Lawler critique applies with even greater force to DSGE and RBC approaches; as James Morley points out, those approaches require the appropriate detrending of the macro series (including GDP). If you know anything about the characteristics of two-sided filters like the HP, Baxter-King BP, etc., and the end-point problem, then you would think twice about the inherent superiority of these approaches vis a vis “old fashioned” models.
By the way, once again at the risk of repeating, if you think all DSGEs of New Keynesian variety imply small multipliers, well, read this post again.
Random Student the kind that got tossed out by macroeconomists three decades ago? (Sims 1980, right?)
Sims’ 1980 criticism of old style structural models was based on “incredible” and ad hoc determinations of exogenous and endogenous variables. His 1980 solution was to develop atheoretical nonstructural VAR models that simply treated all variables as endogenous, so the only theorizing left to the modeler was to determine which variables to include and the number of lags. But you should also mention Sims’ 1986 work in which he solved for structural VARs, which were kind of a compromise between atheoretical VARs and old fashioned structural macro models. Menzie has linked to all kinds of models, ranging from structural models to various VAR models. For example, the CEA & Christine (“Cindy” :->) Romer’s VAR modeling of the effects of ARRA.
And Menzie has linked to his own papers using vectors for I(1) time series with cointegrating vectors. But let’s be clear here, this is a blog intended for folks with an above average interest in economics, but who may not have technical competence in econometric modeling. Appealing to structural macro models have the obvious virtue of being intuitive and accessible while still being useful if done well. And Macroeconomic Advisors does structural modeling quite well.
I’m guessing Mr. Chinn doesn’t like to make things too personal here, but I can’t resist.
Could it be Mr. Lawler is not that good of a writer (usually decent, nevermind great writing requires decent quality thinking, or at least some linear, logical pattern)??? I think Mr. Lawler may enjoy too much his role as ignorant foil, in an attempt to get more visits to his posts at the Spectator. Unfortunately for Mr Lawler, it is not so comedic when there is no “dumb” to go together with his “dumber”.
Not to be confused with Tom Lawler, the housing economist
Your problem, Menzie, is that you expect discourse to respond to facts when in fact you are arguing with belief. Belief is sometimes a matter of filling in a guess to answer a mystery – as in the nature of dark matter – but it is often not only highly resistant to facts that challenge it but bluntly counter-factual. The more facts you present the more it becomes necessary for the believer to deny, distort and outright reject.
We are in some ways as a society re-visiting the trials of Galileo, substituting belief for evidence and treating those who present facts as heretics guilty of assaulting the sacred structure of belief. It is no accident that religion has become associated with politics. For example, they not only deny climate change but now push laws that prevent any regulation of carbon at all, which converts an argument into a theology of belief carved into stone so it resists new facts. It is a closing of the mind because belief trumps facts.
Remember: believers are ruthless because non-believers are heretics.
Reading Lawler’s essay more closely it appears that you cherrypicked certain paragraphs. In fact, he is expressing concerns that are echoed in the comments found in the “Worthwhile Canadian Initiative” you link to.
Specifically, the conclusions that are reached depend on a correctly specified model. If a model fits poorly, that should give you pause about interpreting the conclusions as fact. In fact, you are leaning heavily on the model’s assumptions precisely because the empirics are poor. The poor fit could be due to omitted variable bias, misspecified dynamics, erroneous calibration, etc. To consider the empirics (poor forecast) a misinformed distraction to the issue at hand neglects the fact that the policy conclusion is based upon but one model among many – and that model is pari passu precisely because the empirics have not definitively settled on a single model.
That is the essence of Lawler’s argument and dissenting comments to Stephen Gordon’s (Worthwhile Canadian Initiative) post. A point you don’t seem to ever address. In responding to RandomStudent – you say Lawler resists looking at any counterfactual. That is not true, he is more concerned with how the counterfactual plays out in a multitude of competing models. In this way he is much more in agreement with your post than you give him credit for.
tom: Perhaps I am being too critical, and not allowing for growth, but to me, context is important. Hence, I return to Mr. Lawler’s original exegesis, wherein he writes:
I believe it is legitimate to critique a given model and baseline, be it Troika’s or John Taylor’s. But then Mr. Lawler expands the critique to any baseline.
Those are Mr. Lawler’s own words — not my characterization. I will also note that when I have observed the fact that other researchers (including John Taylor, who has been skeptical of the stimulus) analyze the issue, also use the difference-from-baseline measure, he has not responded.
So, perhaps you see a degree of nuance and subtlety in Mr. Lawler’s writings that I am unable to discern. If that is the case, I am willing to allow that the fault lies with me rather than Mr. Lawler. But I urge you to re-view Mr. Lawler’s earlier posts, and see if you still see the nuance in impute to his writings.
tom I’m not understanding your point. John Lawler devoted a fair amount of “ink” to dissing Romer’s paper, but I have to wonder if he got beyond the unemployment graph and actually read Romer’s analysis. If he read the analysis it should have been pretty clear that the unemployment rate was just a dummied down proxy for the GDP output gap and it should not have been taken quite so literally. Afterall, her analysis was done in late Nov/early Dec 2008, and by that time we now know the unemployment rate was already almost at 8%. But the crux of her argument was that with the stimulus package GDP would begin to turn around in Jun 2009…and that proved to be exactly right. Romer’s paper was really about when GDP would turn around; unfortunately it had to be “sold” in terms of the unemployment rate because that’s what politicians understand. And Romer’s intended audience was not economists; it was policy wonk politicians.
Various models from CBO and private forecasters all came to roughly the same conclusion as the CEA with regard to ARRA’s contribution to GDP growth. Many critics here complain that all of those models shared Keynesian DNA, but so what? Can anyone name some other plausible economic model that would make sense in a ZIRP environment? Austrian mumbo-jumbo??? Real Business Cycle ramblings??? Please. Is the Administration required to take seriously every piece of crackpot economics that Casey Mulligan can imagine? I don’t think so. Obviously the Administration can never “prove” a counterfactual, but the various models that support the Administration’s argument are pretty standard stuff. I don’t think the Administration has any obligation to go beyond standard macroeconomic models. And what model is Lawler appealing to?
There’s a new working paper over at NBER (WP #16759) that looks at the effectiveness of ARRA from a different approach. I’m not entirely comfortable with everything in the paper, but even though the authors are somewhat downbeat about the effectiveness of ARRA, they still conclude:
Our results are somewhat mixed, but generally support the effectiveness of the ARRA. Our point estimates for the stimulus as a whole suggest that it was somewhat less effective than anticipated by the administration, but that their estimates are well within our confidence intervals. Overall we find a cost per job between $100,000 and $400,000 depending on our specification. This implies overall Keynesian multipliers between 0.5 and 2.0. By performing a
state level analysis, we are excluding impacts that cross state lines, which is likely biasing our estimates of the effectiveness down.
The found that monies spent on low income support were especially effective, while spending on education and police did not generate a lot of new jobs. Not really that big of a surprise given that education and public safety spending was more a case of saving jobs rather than creating new jobs.